Invented naming grammar

Invent a system that can explain its output

A repeated decorative ending is not a complete naming culture. Start with the components the grammar may combine, the order of operations and the exceptions your world allows. This tool builds one onset, one nucleus, an optional coda and a fixed suffix. It is a deliberately bounded construction system, not a complete conlang generator or a model of a real language.

Trace one name before generating many

With onset v, nucleus ae, coda n and suffix el, the initial form is vaenel. Each ordered transformation then records a new intermediate form when it changes the string. The trace lets you distinguish an actual derivation from a name that merely looks similar. A fictional gloss belongs in an author note; the tool does not convert invented fragments into real dictionary meanings.

Order and boundaries matter

A transformation applies once at its position in the list, replacing the declared substring wherever it occurs at that stage. It is not restricted to morpheme boundaries. That makes the rule easy to inspect but means a broad substitution can alter a suffix as well as a stem. Review the trace. The tool rejects cyclic transform chains and oversized combination requests rather than running an unbounded rewrite process.

Fictional example: the border interpreter

An interpreter keeps a home lineage component but uses a trade-language rendering in public. A clerk applies the trade rule to every component and changes the lineage marker as well. The correction preserves two linked forms and records which transformation may apply to which use. In a richer production grammar, the boundary rule would need explicit component scope rather than the global substring model used in this compact workbench.

Why a random blend would not explain the scene

If the name came from unrelated attractive syllables, the clerk’s mistake would have no rule the reader could learn. A consistent derivation makes the correction legible. If the home and trade forms were identical, the conflict would need another cause. Design the rule around the story’s actual contact and inheritance decisions instead of assigning one universal style to an entire species.

Test exceptions before enlarging the inventory

Forbidden clusters are checked against the transformed output. Two different derivations may collapse to the same visible name; the result is deduplicated rather than counted twice. If your intended exception is rejected, change the rule deliberately and record why. A larger pool does not repair an unexplained inconsistency. Save the components, transformations, seed and accepted examples so collaborators can reproduce the same decisions.

Declare the relation before deriving a parent-name form

The optional parent worksheet takes child ID, maternal source ID and paternal source ID, with supplied fictional roots in a separate table. A maternal policy cannot borrow the paternal source to fill a missing relation. A root Mira plus an explicitly authored suffix en gives Miraen; this is not a claim about a real language’s matronymics. With neither a transform nor a suffix, derivation stops. Missing sources and cycles also stop commitment. Review the derivation trace in the portable style guide before adopting the output.

Write a grammar someone else can challenge

A naming grammar should explain why a form belongs, not merely produce a large set of strings that look related. Begin with original components and a small number of construction rules. This workbench combines an onset, nucleus, optional coda and suffix, then applies ordered substring transformations. The bounded model makes derivations inspectable. It is not a complete language simulator and should not be described as an account of a real culture.

Supply positive and negative examples before expanding the inventory. A positive example says that the grammar must allow this form; a negative example says it must not. The tool checks those expectations against the full eligible construction set, not merely against the few randomly selected results on screen. Otherwise an invalid positive example could be hidden because it happened not to appear in a batch.

A failed example blocks generation and shows the disagreement. That is a feature. It prevents a collaborator from receiving a polished style guide whose own examples contradict the rule. Change the rule, change an incorrectly specified example or document a genuine exception. Do not delete every difficult case until the test appears to pass.

Reproduce one derivation and one failure

Set the onset to v, nucleus to a, coda to n and suffix to an empty field. With no transforms or forbidden clusters, the construction yields van. Enter van | accept and ven | reject. The examples match the declared inventory. The requested count can still exceed the finite output pool, producing a partial result without invalidating the grammar.

Now change the positive example to ven | accept without adding e to the nucleus inventory. Generation is blocked because the grammar cannot produce the form it is supposed to allow. A lucky reroll cannot fix this. If ven is an intentional personal exception, put it in the exception table with a reason rather than pretend that e was present all along.

Next add a transformation and inspect its trace. Order matters when one rewrite changes the input of the next. Rules apply once in list order, replacing the declared substring at that stage. They are not automatically restricted to a stem boundary. A broad change can affect a suffix too. The workbench rejects cyclic rewrite chains as a conservative guard; it does not offer an unbounded iterative rewrite engine.

Original fictional scene: the apprentice’s exception

The naming exercise was supposed to take ten minutes. The master gave the apprentices three starts, two centers and a shared ending. By noon they had covered the slate with forms that looked like members of one household.

One apprentice wrote a name that did not belong.

“You forgot a rule,” another said.

“I followed the example on the door.”

The master looked at the door. The workshop’s founder had used an older form whose history nobody had put in the lesson. The apprentices had been told to imitate a culture, then penalized for noticing the culture’s actual exception.

The master could have added another vowel and called the problem solved. That would also admit several forms she had meant to exclude. Instead, she wrote two tests: the founder’s name was allowed for that specific identity, while the ordinary pattern remained unchanged.

“Then it is a rule for one person,” the first apprentice said.

“It is a reason we must keep with one person.”

They returned to the slate. The exception did not make every unrelated form valid. It made the style guide more honest about the difference between a living identity and a productive construction.

When a neighboring workshop sent its own naming pattern for a joint event, the apprentices compared the two sets with the same roles and seed. They could explain which components and endings differed instead of deciding that one sounded foreign because its letters were unfamiliar.

This is a fictional teaching scene. The founder and workshops have invented conventions, not attested linguistic histories. Its function is to show why an exception should remain attributed and limited rather than be smuggled into the general rule.

Compare cultures through structure, not decoration

The second-grammar fields let you supply another set of onsets, nuclei, codas and suffix. The comparison uses the same batch conditions so the difference is easier to inspect. It does not establish that every person from one society shares a style or that a biological species has only one culture.

Look beyond the final ending. Which components can appear in which positions? Are names inherited, chosen or earned? The compact generator directly implements only its stated component construction. More elaborate kinship transmission belongs in a declared family rule rather than an explanation invented after a random result. A coherent world can link the tools without pretending that one universal generator models every social process.

If the two grammars produce overlapping visible forms, examine whether that overlap is plausible contact, a deliberate shared tradition or an accidental collision. The comparison table does not automatically label it wrong. Record a decision and test the names in actual references. Distinct structural rules can still produce a pair readers confuse in the same scene.

Exceptions are not a back door around hard filters

An exception needs a name and a reason. It can satisfy a positive grammar example without changing the productive inventory. The ordinary exclusion and saved-name rules still constrain which results can appear in the output. An author may approve an irregular world form while choosing not to generate it again in this project.

Two different derivations can collapse to the same visible name after transformations. The tool deduplicates them rather than counting them as two independent candidates. Preserve a useful derivation trace, and do not advertise the nominal product of inventory sizes as the number of distinct usable results after all filters. A finite pool report should describe the eligible result state, not an impressive prefilter total.

When a prohibited cluster appears only after rewriting, the transformed form must be rejected. Checking components independently would miss that boundary effect. Conversely, an authored exception should state exactly what it overrides. If your intended exception does not fit the compact model, describe the gap instead of claiming the workbench silently handled a more sophisticated phonology.

Produce a portable style guide

The export records the authored components, transforms, examples, tests, exceptions and version. Its fictional-gloss label prevents an assigned story meaning from becoming a dictionary assertion. Keep that version with any accepted character identities so later collaborators can distinguish an older spelling from an unexplained inconsistency.

Before changing the grammar, run the existing positive and negative examples. If a reform is an intentional world event, preserve earlier forms and specify where the new rule applies. A spelling reform, a dialect adaptation and a personal rename are different story operations. They may share data, but they should not erase the same history without explanation.

Read a small cast generated by the grammar in prose and aloud using approved readings. A technically valid pattern may be tiring when every character has the same rhythm. That observation can motivate a change in the allowed inventory or reference policy, but it does not license fabricated real-world etymologies. The strongest output is a set of names whose derivation, exceptions and narrative use another creator can actually understand.

Why does a positive example stop generation?

The grammar cannot honestly claim to satisfy an example it cannot produce. The test checks the eligible construction set, not only the sampled batch. Repair the inventory or transform, correct the example, or document an actual exception. Deleting the test merely to make generation resume hides the inconsistency from the next collaborator and turns the style guide into a collection of unsupported impressions.

Can an exception override my exclusion list?

No. A fictional irregular form can be valid in the world while remaining excluded from this project’s new candidates. The exception needs an authored reason and is still subject to the output’s hard exclusion and saved-name rules. This separation prevents a useful historical or personal exception from becoming a back door that silently breaks a user’s generation constraints.

Do the transforms operate only on the stem?

No. This compact model applies ordered substring replacements to the current full form. Inspect the intermediate trace because a broad pattern can also affect an ending. Cyclic chains are conservatively rejected. If your intended phonology needs richer boundary, stress or morphology rules, record that requirement explicitly rather than describe the current output as though those unimplemented rules had already run.

What does comparing two grammars prove?

It shows outputs under two supplied component systems using controlled batch conditions. It does not prove cultural authenticity, linguistic naturalness or perceptual distinction. Use the comparison to explain structural choices, then check the actual cast. Different grammars can produce overlapping forms, and one culture can contain several naming traditions. Do not turn the comparison into a one-species, one-style assumption.